Imported from oimiragieo/agent-studio (.claude/skills/content-analyzer/SKILL.md). Install upstream with npx skills add oimiragieo/agent-studio --skill content-analyzer. Copyright stays with the author.

Content Analyzer

Overview

Multi-dimensional content analysis skill that transforms published text into actionable engagement intelligence. The analysis pipeline covers six dimensions: sentiment detection, readability scoring, structural analysis, hook classification, topic identification, and engagement correlation.

Core principle: Content quality is measurable. Every post has structural attributes that correlate with engagement outcomes. This skill quantifies those attributes and tracks them over time.

When to Invoke

Skill({ skill: 'content-analyzer' });

Invoke when:

• Analyzing a published blog post, article, or social media post
• Running a daily content performance audit
• Identifying what makes top-performing content succeed
• Building a content strategy based on historical performance data
• Diagnosing why a post underperformed expectations
• Comparing content attributes across a portfolio of published work

Do NOT invoke for:

• Pre-publication copyediting (use a writing or editing skill instead)
• SEO keyword research without existing content (use seo-optimization)
• Social media scheduling or publishing automation

Six-Dimension Analysis Pipeline

Dimension 1: Sentiment Analysis

Classify the emotional tone of the content across three axes:

Polarity: Positive / Negative / Neutral / Mixed Emotion: Inspiration, Urgency, Curiosity, Empathy, Authority, Humor, Fear Intensity: 1 (subtle) to 5 (intense)

Scoring method:

• Read the full text and identify the dominant emotional register
• Check for tonal shifts (e.g., problem-heavy opening -> optimistic conclusion)
• Score intensity based on word choice strength and punctuation patterns
• Flag mixed sentiment (common in "problem-solution" posts)

Output:

{
  "sentiment": {
    "polarity": "positive",
    "dominantEmotion": "curiosity",
    "intensity": 3,
    "tonalShifts": ["negative->positive at paragraph 4"],
    "emotionBreakdown": {
      "curiosity": 0.45,
      "authority": 0.3,
      "urgency": 0.15,
      "empathy": 0.1
    }
  }
}

Dimension 2: Readability Scoring

Quantify how easy the content is to read and understand:

Output:

{
  "readability": {
    "fleschKincaidGrade": 7.2,
    "avgSentenceLength": 16.4,
    "vocabularyComplexity": 0.11,
    "avgParagraphLength": 58,
    "passiveVoicePercent": 0.06,
    "rating": "GOOD"
  }
}

Dimension 3: Structural Analysis

Evaluate the architectural quality of the content:

Elements to analyze:

• Opening hook: First 1-2 sentences -- what technique is used?
• Section structure: Number of headings, heading hierarchy, section balance
• Visual breaks: Lists, blockquotes, images, code blocks, callouts
• CTA presence: Where CTAs appear, how strong they are, how many
• Closing: How the post ends (summary, CTA, question, cliffhanger)

Hook Classification Taxonomy:

Output:

{
  "structure": {
    "hookType": "statistic",
    "hookText": "78% of content marketers...",
    "headingCount": 6,
    "headingHierarchy": "h1->h2->h3 (consistent)",
    "visualBreaks": { "lists": 4, "blockquotes": 1, "images": 2, "codeBlocks": 0 },
    "ctaCount": 2,
    "ctaPositions": ["mid-article", "closing"],
    "ctaStrength": "medium",
    "closingType": "question",
    "wordCount": 1847,
    "estimatedReadTime": "8 min"
  }
}

Dimension 4: Topic Classification

Identify what the content is about and how it fits into topic clusters:

• Primary topic: The main subject (1 topic)
• Secondary topics: Related themes (2-3 topics)
• Keyword density: Top 10 keywords with frequency
• Topic cluster: Which content pillar this belongs to

Output:

{
  "topics": {
    "primary": "content marketing strategy",
    "secondary": ["SEO optimization", "audience engagement"],
    "topKeywords": [
      { "keyword": "content", "count": 24, "density": 0.013 },
      { "keyword": "engagement", "count": 18, "density": 0.01 }
    ],
    "cluster": "content-strategy"
  }
}

Dimension 5: Wording Pattern Analysis

Analyze the specific language choices that drive engagement:

• Power words: Words that trigger emotion (e.g., "discover", "secret", "proven")
• Transition words: Connectors that improve flow (e.g., "however", "specifically")
• Jargon level: Domain-specific terms as % of total vocabulary
• Personal pronouns: "you/your" frequency (reader-focus indicator)
• Action verbs: Active vs passive construction ratio

Output:

{
  "wording": {
    "powerWordCount": 14,
    "powerWordDensity": 0.008,
    "transitionWordCount": 22,
    "jargonLevel": "low",
    "personalPronounDensity": 0.032,
    "actionVerbRatio": 0.78,
    "topPowerWords": ["discover", "proven", "essential", "transform"]
  }
}

Dimension 6: Engagement Correlation

Map content attributes to engagement outcomes (requires engagement data):

Correlation analysis:

For each content attribute (hook type, length, sentiment, topic):
  1. Group posts by attribute value
  2. Calculate average engagement per group
  3. Rank groups by engagement
  4. Identify statistically significant differences

Output:

{
  "engagement": {
    "metrics": {
      "views": 4521,
      "likes": 234,
      "comments": 47,
      "shares": 89,
      "bookmarks": 156,
      "engagementRate": 0.116
    },
    "correlations": {
      "hookType": { "statistic": 1.4, "question": 1.2, "story": 1.0, "howTo": 0.8 },
      "lengthBucket": { "1500-2000": 1.3, "1000-1500": 1.1, "2000+": 0.9, "<1000": 0.7 },
      "sentimentTone": { "curiosity": 1.5, "authority": 1.2, "urgency": 0.9 }
    },
    "confidenceLevel": "medium",
    "sampleSize": 23
  }
}

CLI Integration

Run the analysis via the post-analyzer CLI tool:

# Analyze a single URL
node .claude/tools/cli/post-analyzer.cjs --url "https://example.com/post" --output json

# Analyze from local text file
node .claude/tools/cli/post-analyzer.cjs --file "./content.txt" --output json

# Generate daily report
node .claude/tools/cli/post-analyzer.cjs --url "https://example.com/post" --report daily

Report Generation

After analysis, generate a daily report using the template:

.claude/templates/reports/daily-content-report.md

Output to:

.claude/context/reports/backend/daily-content-report-{YYYY-MM-DD}.md

Historical Data Storage

All analysis results are appended to:

.claude/context/data/content-analytics.json

Schema:

{
  "analyses": [
    {
      "id": "analysis-{timestamp}",
      "url": "https://example.com/post",
      "analyzedAt": "2026-03-21T10:00:00Z",
      "title": "Post Title",
      "sentiment": {},
      "readability": {},
      "structure": {},
      "topics": {},
      "wording": {},
      "engagement": {}
    }
  ],
  "trends": {
    "7day": {},
    "30day": {}
  },
  "lastUpdated": "2026-03-21T10:00:00Z"
}

Iron Laws

1. ALWAYS analyze all six dimensions for every post -- partial analysis produces misleading conclusions.
2. ALWAYS store results in content-analytics.json after every analysis -- trend detection requires complete historical data.
3. NEVER report engagement correlations from fewer than 5 data points -- small samples produce spurious patterns.
4. ALWAYS classify the hook type before reporting engagement drivers -- the hook is the strongest single predictor of click-through performance.
5. NEVER scrape content without respecting rate limits and robots.txt -- getting blocked destroys the analysis pipeline.

Anti-Patterns

Assigned Agents

This skill is used by:

• post-analyzer-agent -- Primary: daily content analysis and reporting
• feedback-synthesizer -- Supporting: when content feedback intersects with customer feedback
• researcher -- Supporting: when content research needs quantitative analysis

Memory Protocol (MANDATORY)

Before starting:

node .claude/lib/memory/memory-search.cjs "content analysis engagement hooks sentiment"

Read .claude/context/memory/learnings.md

Check for:

• Previous content analysis results and patterns
• Known engagement driver correlations
• Historical trend data in content-analytics.json

After completing:

• New content pattern discovered -> .claude/context/memory/learnings.md
• Analysis pipeline issue -> .claude/context/memory/issues.md
• Engagement correlation decision -> .claude/context/memory/decisions.md

ASSUME INTERRUPTION: Your context may reset. If it's not in memory, it didn't happen.